Aalborg Universitet
System Level Analysis of eMBB and Grant-Free URLLC Multiplexing in Uplink
Abreu, Renato Barbosa; Jacobsen, Thomas; Pedersen, Klaus I.; Berardinelli, Gilberto;
Mogensen, Preben Elgaard
Published in:
2019 IEEE 89th Vehicular Technology Conference (VTC Spring)
DOI (link to publication from Publisher):
10.1109/VTCSpring.2019.8746557
Publication date:
2019
Document Version
Accepted author manuscript, peer reviewed version
Link to publication from Aalborg University
Citation for published version (APA):
Abreu, R. B., Jacobsen, T., Pedersen, K. I., Berardinelli, G., & Mogensen, P. E. (2019). System Level Analysis
of eMBB and Grant-Free URLLC Multiplexing in Uplink. In 2019 IEEE 89th Vehicular Technology Conference
(VTC Spring) [8746557] IEEE. I E E E V T S Vehicular Technology Conference. Proceedings
https://doi.org/10.1109/VTCSpring.2019.8746557
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
- Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
- You may not further distribute the material or use it for any profit-making activity or commercial gain
- You may freely distribute the URL identifying the publication in the public portal -
Take down policy
If you believe that this document breaches copyright please contact us at vbn@aub.aau.dk providing details, and we will remove access to
the work immediately and investigate your claim.
Downloaded from vbn.aau.dk on: August 10, 2022
1
System Level Analysis of eMBB and Grant-Free
URLLC Multiplexing in Uplink
Renato Abreu
∗
, Thoma s Jacobsen
∗
, Klaus Pedersen
∗†
, Gilberto Berardinelli
∗
, Preben Mo gensen
∗†
∗
Dept. of Electronic Systems, Aalborg University;
†
Nokia Bell Labs, Aalborg, Denmark
Email: {rba, tj, kip, gb, pm}@es.aau.dk
Abstract—5th generation radio networks should efficiently
support services with diverse requirements. For achieving better
resource utilization, the sharing of the radio channel between th e
different services is an attractive solution. While the downlink
multiplexing can be well accomplished with dynamic scheduling,
efficient multiplexing of enhan ced mobile broadband (eMBB) and
ultra-reliable low-latency communications (URLLC) in uplink is
still an open problem. In particular, we consider the case of
URLLC using grant-free allocation for sporadic transmissions,
multiplexed on shared resources with eMBB with high data
volume. Since the moment in whi ch a grant-free transmission
occurs is not known, URL LC and eMBB transmissions overlay.
Power control settings are then assessed as a way to manage
the performance trade-off between the services. Due to t he
complexity of 5G NR, the evaluation is based on advanced
system level simulations. Insights regarding the configuration
of fractional power control settings up on the coexistence of the
different services are presented.
I. INTRODUCTION
The recent 5th gene ration (5G) new radio (NR) specifi-
cations include features for conveying traffic with different
characteristics and requirements. One example is enhanced
mobile broadband (eMBB) which focuses on high volume of
data transmissions, demanding high spectral efficiency. Ultra-
reliable low-latency communications (URLLC) target instead,
to de liver intermittent small payloads with high success prob-
ability in a short time interval. A baseline target for URLLC
is to enable transmissions over the air interface o f 32 bytes
payload s within 1 ms and a 1−10
−5
reliability [1]. The initial
support of each of the se services is readily provided by the
3GPP Release-15 specification [2]. However, the multiplexing
of u plink traffic with different reliability r equirements has
gained attention, given the need of supporting heterogeneous
services while ensuring efficient use of the radio resources [3].
The efficient multiplexing of eMBB and URLLC in down-
link can b e achieved by dynamic scheduling , with th e high
priority URLLC transmissions punc turing the eMBB alloca-
tion [4]. In uplink, sim ilar concept c an be employed with
preemption schemes, both for intra-UE (for the same UE) an d
for inter-UE (between different UEs) tra ffic multiplexing. With
this, e MBB transmission is paused w hile URLLC is granted
to transmit. While this solution is valid f or dynamic scheduled
transmissions, the same is no t applicable when grant-free
schemes are utilized. Grant-free transmissions, specified as
configured grants in NR [5], is one of the main enablers of
uplink URLLC with very stringent requirements. In that, the
resource allocation, as well as other physical layer parameter s,
are pre-configured by radio resource control (RRC) signaling.
Thus, the usual handshake process, of sending a scheduling
request and waiting for a grant for every transmission, can
be avoided. This reduces not only the delay, but also the
dependence of error-prone co ntrol signaling for URLLC. For
reducing the resource wastage caused by sporadic URLLC
transmissions, the base station (BS) can config ure the sam e
resources to multiple user equipments (UE). However, this
leads to augmented intra-cell interference when transmissions
overlap. The problem becomes more evident if the grant-
free resourc es are overlaid for multiplexing abundant eMBB
traffic. Since it is not known a priori whe n a sporadic URLLC
transmission will occur, it is not possible to timely interr upt
an on going transmission for avoiding a collision, potentially
degrading the reliability.
Different studies have considered the problem of multi-
plexing heterogeneous traffic in uplink. In [6], a joint eMBB
and URLLC scheduler is prop osed, with superposition of
ongoing transmissions. The overlayin g multiplexing between
resource greedy broadband tra ffic and sporadic small data
is considered in [7] and evaluated with basic informatio n
theoretical tools for a single cell scenario. An heterog eneous
non-orthogonal multiple access approach is studied in [8]
using a theoretic model, however, multiple URLLC transmis-
sions over the share d resou rce are not c onsidered. In [9], a
theoretical analysis of overlaying versus separate allo cation is
presented. Minimum-mean square error (MMSE) is considered
for th e reception of multiple URLL C and eMBB transmissions.
Detailed analysis considering the aspects of a multi-cell 5G
NR system are not considered in previous work s.
In this work we present system level performanc e evaluation
for the inter-UE multiplexing of e MBB and URLLC uplink
transmissions. We consider the c a se of sporadic grant-free
URLLC, with shared resource allocations, overlayin g with
full-buffer eMBB stre ams, in a multi-cell system. We discuss
the aspects of op e n loop power control and identify the criteria
for setting the relevant parameters in order to manage the
trade-off between URLLC reliability and eMBB capacity.
Results from d etailed simulation campaigns following 5G
NR a ssumptions are presented in terms of URLL C outage
probability an d eMBB SINR.
The reminder of the work is organized as follows. The
considered system is presented in Section II and the power
control aspects in Sectio n III. Section I V de scribes the m ethod-
ology and assumption s. Results are presented in Section V and
discussed in Section VI. Section VII concludes the paper.
2
W
mini-slot size T
time
freq.
N
e
simultaneous eMBB streams
up to N
u
sporadic URLLC transmissions
Fig. 1. Overlaying eMBB and grant-free URLLC allocations in a cell.
II. SYSTEM MODEL
We consider a multi-cell r a dio network composed of C
cells with sync hronized base stations (BS). A fixed number of
URLLC UEs N
u
are dep loyed in each cell. Besides, N
e
eMBB
UEs can be active in the same cell. The UEs are considered
to be connected and synchro nized with th e serv ing BS for
their uplink data transm ission. Fig. 1 illustrates the con sidered
multiplexing scheme. The eMBB UEs are assumed to have
a large amount of data to transmit. Their traffic follows a
full buffer m odel, ensuring a permanent flow of eMBB data
to be scheduled over the time slots. The N
e
eMBB UEs
are sche duled over the full carrier bandw idth W . The BS
exploits then multi-user rec eption capability by employing an
M
r
antennas receiver, for retrieving overlaying signals.
The URLLC UEs have sporadic traffic consisted of small
payload s of size B. Such traffic is modele d as a Poisson
arrival process with packet arrival rate λ. In order to serve the
URLLC tr affic with minimum latency, a short-TTI of duration
T is employed. The serving BS configu res also th e URLLC
UEs to transmit with grant-fr ee resources over the bandwidth
W . We assume that the N
u
UEs share the same resource
configuration, therefore their transmissions are susceptible to
mutual collisions, in addition to the in te rference fr om eMBB
traffic being multiplexed over the same resources. A wide-
band allocation allows harvesting frequency diversity. It also
permits the use of a robust modulation and c oding scheme
(MCS) to cope with fading and potential interference from
simultaneou s tran sm issions.
A linear minimum-mean square error with interference
rejection c ombining (MMSE-IRC) receiver is assumed in the
BS. Since the UEs and the BSs are fully synchronized, it
permits the receiver to take into account in tra- and inter-cell
interference signa ls f or computing the interference covariance
matrix. Then, the MMSE-IRC r eceiver op erates on the degrees
of freedom offered by the multiple receive antennas to retrieve
multiple overlaid transmissions. Still, in case the interference
level is too severe the re ception can be compromised. This
motivates the use of careful power control settings for red ucing
the pen alty in the URLLC reliability o r eMBB capacity.
III. POWER CONTROL SETTING FOR OVERLAYING
TRANSMISSIONS
The 3GPP Release-15 specification defines the power con-
trol for the uplink channels in [10]. The transmit power (in
dBm) over the physical uplink shared channel (PUSCH) is
described, in simplified notation, as
P = min
(
P
max
P
0
+ 10log
10
(2
µ
M) + αP L + ∆
mcs
+ f (i)
,
(1)
where P
max
is the maximum transmit power of the UE, P
0
is a UE specific parameter related to the power per resourc e
block (RB), the exponent µ is set according the sub-carrier
spacing (0 for 15 kHz, 1 for 30 kHz, and so on), M is the
number of RBs a llocated, α is a path-loss compensation factor,
P L is the estimated p ath-loss between the UE and the BS.
∆
MCS
is a quality requireme nt parameter depending on the
MCS th a t can be configured by upper layers and f (i) is a
parameter for closed lo op power control adjustments; these
were not considered in this study.
The use of fractional power control is known f or improving
the capacity for broadband co mmunicatio n [11]. For such,
α < 1 is a pplied, as well as a correspondent increase in P
0
,
improving the SINR, and h ence, the throughput of c ell center
UEs. However, as discussed in [12] , the usage of full p ath-loss
compen sation is more attractive for URLLC to avoid an outage
penalty in cell edge . In the case o f overlaying allocations, the
performance of eMBB and URLLC presents a trade-off, i.e.
power control settings that benefits eMBB pen alizes URLLC
and vice-versa. Thus, in our proposal the settings are applied
on a service basis. With tha t, eMBB UEs are configured
with P
e
0
and α
e
, while URLLC UEs ar e configured with P
u
0
and α
u
. Here we assume that, for each service, all UEs in
the cell use the same parameters. These parameters should
be carefully selected fo r mee ting the service require ments.
As a simple example, for α
u
= α
e
setting P
e
0
>> P
u
0
potentially inc reases the interf erence of eMBB over URLLC
compromising the reliability. While P
e
0
<< P
u
0
can deteriorate
the eMBB capac ity.
IV. EVALUATION METHODOLOGY
The impact on the performance of overlaying grant-free
URLLC and eMBB is evaluated through extensive system
level simulations for different power control settings. The
evaluation methodology is based on NR assumptions as de-
fined in [13]. The simulator uses commonly accep te d models
and is calibrated accor ding to 3GPP NR guidelines [14]. The
main parameters for the network configuration and the main
simulation assumptio ns are summarized in Table I.
A 3D urban macro scenario is assumed, consisting of
C = 21 syn chronized cells (7 sites with 3 sectors each).
The inter-site distan ce is 500 meters. World wrap around is
used for avoiding edge effects. We consider different load
conditions for URLLC. For low load, 10 URLLC UEs per
cell are uniformly distributed in the scenario. And for high
load, 300 URLLC UEs per cell are distributed. Each URLLC
UE transmits payloads of B = 32 bytes following a Poisson
arrival process with average arrival interval of 100 ms, i.e.
λ = 10 packets per second. This leads to a load L = 25.6 kbps
per c ell for low URLLC load, and L = 768 kbps for high
URLLC load. One and two eMBB UEs are also deployed in
each cell, equivalent to a single stream and two multi-user
MIMO streams. The eMBB UEs use f ull-buffer traffic model,
3
TABLE I
SI MULATION AS SUMP TIONS
Parameters Assumption
Layout Hexagonal grid with 21 cells (7 sites and
3 sectors/site), world wrap-around
Inter-site distance 500 meters
Carrier frequency 4 GHz
Channel model 3D Urban Macro (UMa)
UE distribution Uniformly distributed outdoor, 3 km/h UE
speed fading model
UE transmitter P
max
= 23 dBm, M
t
= 1 transmit antenna
BS receiver MMSE-IRC, M
r
= 4 receive antennas
Receiver noise figure 5 dB
Thermal noise −174 dBm/Hz
Bandwidth W = 10 MHz in uplink, FDD
PHY configuration 15 kHz sub-carrier spacing, 2 symbols mini-
slot (T = 0.143 ms), 12 sub-carriers/RB
Grant-free configura-
tion
MCS QPSK1/8, periodicity of 2 symbols,
M = 48 RBs for uplink data, HARQ disabled
eMBB UEs per cell 0 (no eMBB interference baseline), 1 (single
stream) and 2 (MU-MIMO streams)
eMBB traffic model full-buffer
URLLC UEs per cell 10 for low load, and 300 for high load
URLLC traffic model FTP Model 3, B = 32 bytes, Poisson arrival
rate of λ = 10 packets per second per UE
being continuously scheduled over the full bandwidth. The
UEs are deployed at the beginning of the simulation drop.
Each UE connects to the cell with highest reference sign a l
received power (RSRP) and remains in connected state until
the simulation finishes.
The URLLC UEs are configured for transmission in mini-
slots of 2 OFDM symbols, with sub-carrier spacing of 15 kHz
which leads to a T = 0.143 ms TTI. The allocation for grant-
free transmissions uses a bandwidth W = 10 MHz, giving
M = 48 RBs for data, with 2 symbols periodicity. T his
allows a tr ansmission opportunity in full-band at every TTI
in order to minimize latency. The grant-f ree transmissions use
a conservative MCS QPSK 1/8, fitting the 32 bytes payload
in one- shot transmission without segmentation. Considering
latest proc essing time assumptions (capability 2 in [10]),
a transmission can be received and processed within 1 ms.
HARQ retransmissions ar e not considered.
The BSs are equipped with MMSE-IRC with M
r
= 4
receive antennas. Channel estimation is assumed ide al for
the desired and interference signals. The successful recep-
tion of a packet depends on the obtained post-proc e ssing
SINR at the receiver and the used MCS. For every detected
transmission, the post-pro cessing SINR after the MMSE-IRC
receiver combining is calculated for each su b-carrier. That is
used to compute the symbo l-level mutu al information metric
accordin g to the applied modulation as describ e d in [15].
Then, given the used code rate, a look- up table obtained from
extensive link level simulations is used to map the metric value
to a block e rror probability.
Multiple simulation drops a re executed for collecting 5
million URLLC transmission samples, in or der to obtain
statistically significant results in the low qua ntiles [16]. The
main key performance indicator analyzed for URLLC is the
outage probability, i.e. the complemen t of the reliability (tar-
-130 -120 -110 -100 -90 -80 -70
Coupling gain [dB]
0
0.2
0.4
0.6
0.8
1
CDF
-20 0 20
URLLC transmit power [dBm]
0
0.2
0.4
0.6
0.8
1
CDF
-20 0 20
eMBB transmit power [dBm]
0
0.2
0.4
0.6
0.8
1
CDF
Fig. 2. Coupling gain distribution in evaluated urban macro scenario outdoor
(top). Transmit power distribution for URLLC UEs (bottom left), and eMBB
UEs (bottom right).
geting 10
−5
). The latency of each tra nsmission is used for
determining an empirical complementary cumulative distribu-
tion functions (CCDF). The outage probability is then read at
the 1 ms from the latency CCDF. For the eMBB performanc e ,
we collect the 5th percentile and the 50th percentile SINR
values. These reference metrics indicate the cell edge an d the
near to average pe rformance, respectively.
V. PERFORMANCE EVALUATION
The power control settings P
0
and α for e MBB and U RLLC
UEs were varied for the different simulation campaigns, in
which were collec ted the one-way latency of the URLLC
packets a nd the SINR of the eMBB transmissions. The power
control settings for URLLC were chosen as the on es that allow
the highest URLLC load while fulfilling the requ irements [12].
Full path-loss compensation is u sed for URLLC, i.e. α
u
= 1.
For eMBB, full and fractional path- loss compe nsation a re
used, i.e. α
e
= 1 and α
e
= 0.7 respectively. The P
0
values ar e set equal or lower than the URLLC ones, except
when fractiona l path-loss compensation is used. For re ference,
the empirical cumulative distribution function (CDF) of the
coupling gain for the evaluated outdoo r scenario is shown in
Fig. 2. The CDFs of the URLLC and the eM BB transmit power
are also shown for each utilized setting. For both, URLLC and
eMBB using α
u
= α
e
= 1 and P
u
0
= P
e
0
= −108 dBm, 3%
of the UEs transmit with maximum power P
max
. For URLLC
configured with conservative power control settings, α
u
= 1
and P
u
0
= −103 dBm, 15% of the URLLC UEs tra nsmit with
P
max
. For eMBB with α
e
= 0.7 and P
e
0
= −78 dBm, as well
as with α
e
= 1 and P
e
0
= −113 dBm, virtually no eMBB UE
reaches P
max
.
Fig.3 shows the outage probability for the case of 10
URLLC UEs per cell, with th eir transmissions being multi-
plexed with 1 an d with 2 eMBB interferer streams. Baseline
cases without eMBB interference are also shown as “eMBB
off”. It is observed that th e URLLC target is satisfied if
no eMBB UEs are present, leading to an outage pr obability
< 10
−6
. Reducing the power of eMBB with P
e
0
= −113 dBm
4
eMBB
off
=1.0,
P
=-113
=1.0,
P
=-108
=0.7,
P
=-78
eMBB
off
=1.0,
P
=-113
=1.0,
P
=-108
=0.7,
P
=-78
URLLC P
=-108
1,00E-06 2,50E-06 5,26E-05 1,21E-04 1,00E-06 6,35E-05 1,18E-03 6,32E-03
URLLC P
=-103
1,00E-06 1,00E-06 2,55E-05 2,45E-05 1,00E-06 2,58E-05 2,55E-04 2,96E-04
1,00E-06
1,00E-05
1,00E-04
1,00E-03
1,00E-02
URLLC outage probability
1 eMBB interferer stream 2 eMBB interferer stream
u
u
e
e
e
e
e
e
e
e
e
e
e
e
Fig. 3. Outage probability of grant-free URLLC for L = 25.6 kbps.
eMBB
off
α =1.0,
P₀=-113
α =1.0,
P₀=-108
α =0.7,
P₀=-78
eMBB
off
α =1.0,
P₀=-113
α =1.0,
P₀=-108
α =0.7,
P₀=-78
URLLC P₀=-108
1,25E-05 9,96E-05 4,37E-04 1,57E-03 1,25E-05 3,60E-04 3,85E-03 1,16E-02
URLLC P₀=-103
2,65E-05 7,96E-05 2,56E-04 2,33E-04 2,65E-05 1,94E-04 1,14E-03 1,17E-03
1,00E-06
1,00E-05
1,00E-04
1,00E-03
1,00E-02
URLLC outage probability
1 eMBB interferer stream 2 eMBB interferer stream
u
u
e
e
e
e
e
e
e
e
e
e
e
e
Fig. 4. Outage probability of grant-free URLLC for L = 768 kbps.
(i.e. 5 dB lower than for the URLLC UE) a lso allows URLLC
to rea ch the target, when only 1 eM BB stream is present. For
the cases where eMBB uses the same power control settings
as URLLC, the outage probability rises to the order of 10
−4
.
With 2 simultaneous eMBB streams, the penalty f or URLLC
is obviously higher due to the increased interference. The use
of fractional path-loss compensation for eMBB does not help,
since the cell center eMBB UEs generates higher intra-cell
interference. The outage probability for high URLLC load,
with 300 URLLC UEs per cell, is shown in Fig.4. In this
case the URLL C r equirement is nearly met only whe n eMBB
UEs are not transmitting, i.e . without eMBB in te rference a
URLLC load of ≈ 0.77 Mbps per cell is supported . However,
the outage probability of URLLC increases b y a factor of 10
to 100 when eMBB is present. For both load situations, the use
of a high P
u
0
makes URLLC more robust to the presence of
eMBB interference. However, when eMBB is not present, the
lower P
u
0
results in a lower outage due to reduced interference
among URLLC UEs. Using lower P
e
0
values redu ces the
impact on URLLC, however it comes with the cost of lower
SINR for eMBB, which converts to a capacity loss.
Fig.5 and Fig.6 shows the impac t on the eMBB SINR for the
different power control settings. For the lower URLLC load
there is little difference on eMBB perform ance for the different
URLLC P
u
0
settings. As expected, the eMBB SI N R is low in
the case of a low P
e
0
. And from full to fractional path-loss
compen sation, there is an improvement in the 50th percentile
SINR and a degradation in the 5th percentile SINR. The same
observation can be drawn for one and for two eMBB streams.
With the high er URLLC load ther e is a clear impact in the
eMBB SINR (up to 3.1 dB for P
u
0
= −108 dBm). Besides,
the 5 dB increase in P
u
0
, causes up to 1.6 7 dB of degra dation
in eMBB SINR. The low 5th percentile SINR values, getting
down to −5 dB, indicates the very limited eMBB capacity in
the cell edge even with high P
e
0
.
It is worth to mention that the resource utilization withou t
eMBB, for low URLLC load is 1.4%, an d for high URLLC
load is 35%. This means that a big share of the resources
is wasted in detriment of URLLC. This dem onstrates the
importance of multiplexing eMBB toge ther with the URLLC
traffic for the feasibility of the 5 G system.
VI. DISCUSSION
It is worth noting that, despite the potential of fractional
path-loss compensation for improving eMBB average through-
put, cell center eMBB UEs with elevated transmit power
further penalizes the URLLC transmissions. Therefore, full
path-loss compensation and lower P
0
values should be also
preferred for eMBB when multiplexing with URLLC.
The presence of a high URLLC lo ad in the cell imposes a
reduced capacity for eMBB. The use o f the receiver capability
for MU-MIMO is compromised due to the limitation on
degrees of freedom for suppressing all the m utual interference.
The sy stem performance can be en hanced e.g., by utilizing
MMSE-IRC with higher n umber o f anten nas, which im-
proves the diversity order and interference rejection capability.
Besides, successive interference cancellation (SIC) can be
employed for subtracting the sig nal from decode d URLLC
transmissions from the received signal. This can mainly r e duce
the interference over the eMBB transmissions [8], [9].
For applications in which the latency requirem ent can be
relaxed, preemption schemes enabled by dynamic downlink
control signal should be preferred [17]. T hose are able to in-
terrupt on-going eMBB transmissions for scheduling URLLC
data. eMBB can be potentially resumed after the URLLC
transmission. With that, both U RLLC and eMBB should be
benefited from the reduced in terference. Besides, dynamic
scheduling permits accurate resource allocation and a daptation
per-user transmission ba sis. This re sults in guaranteed quality
of service with efficient usage of resources.
VII. CONCLUSIONS
In this paper, we stud ie d the performance of gr a nt-free
URLLC and eMBB multiplexing in uplink. We consider ed
the overlaying of eMBB transmissions with the grant-free
URLLC transmissions over the same resources. Different up-
link transmit power control settings are proposed for mana ging
the trade-off between the URLLC outage probability and the
eMBB capacity. Detailed evaluation of the settings was con-
ducted through extensive system level simulations following
5G NR assumptions. We observe that overlaying URLLC and
eMBB transmissions is only feasible for low URLLC loa ds
(e.g. 0.26 Mbps). Even though, it requires restrictions which
impose severe performance loss for eMBB, such as, reduced
capability for co-scheduling users and 5 dB lower P
0
value.
